Will AI Take My Job?

The base score is not ours. It is the automation exposure score published by Jobs and Skills Australia in its Generative AI Capacity Study, which scored the task content of every ANZSCO occupation for how much generative AI could undertake. We use it for 357 four-digit occupation groups, joined to JSA's own employment projections for 2025 to 2035.

Exposure is not displacement. This is the single most important thing on the page. An exposure score describes what the technology could do with an occupation's tasks — not whether any employer will act on it, when, or to whom. JSA found 79% of Australian workers are in occupations at low automation risk. The Department of Employment and Workplace Relations reported in July 2026 that the most-exposed occupations were still growing. The International Labour Organization warned in February 2026 that exposure indicators assume static tasks and ignore the adoption gap. MIT researchers found in 2024 that only around 23% of the wages tied to exposure-eligible tasks were economically worth automating at then-current costs.

Bands are calibrated, not invented. The Low, Moderate, High and Very High cut-offs are employment-weighted percentiles of the JSA score across the Australian workforce, so that the Low band covers roughly the same 79% of workers JSA describes as low risk. Round-number cut-offs would have placed three-quarters of the workforce at moderate or worse, which would contradict the source this tool is built on.

What the evidence establishes, and what is our judgment. The direction of each adjustment is supported by published research: social and relationship intensive work has been the most durable against automation (Deming, 2017); physical work in unpredictable settings remains the hardest to automate because dexterity, not software, is the bottleneck (Frey and Osborne's perception-and-manipulation bottleneck); offshorability is a distinct and separately measurable exposure (Blinder). The size of each adjustment is our editorial judgment, not a derived quantity, and we would rather say so than imply a precision the research does not support. Your answers are also capped relative to your occupation's own score, so personal responses adjust the published data rather than overwhelm it.

The 2031 figure is a scenario, not a forecast. Its direction comes from the measured trend in how long a task AI systems can complete autonomously, which has been doubling on a scale of months, and from published projections that humanoid robotics deployment stays concentrated in structured industrial settings until well after 2030. No published method converts those trends into a number of points on an occupation exposure scale — the size of the escalation is our assumption. Nobody has measured 2031. The employer-type adjustment reduces that projected escalation for public-sector workers on the basis that institutional adoption tends to be slower and consultation obligations stronger; it is an assumption rather than a measurement, and it deliberately does not change your exposure today, because the same tasks are exposed regardless of who employs you.

What this must not be used for. This is general information for personal career planning. It is not evidence about any specific role or any specific person, and it should not be used to justify a redundancy. Whether a dismissal is a genuine redundancy is decided under s.389 of the Fair Work Act — on whether the job is still required, whether consultation obligations were met and whether redeployment was reasonable — never by an exposure score.

Source: Jobs and Skills Australia, Generative AI Capacity Study (occupation data on AI exposure), CC BY 4.0 (c) Commonwealth of Australia, occupation table last modified 2025-09-30. Employment projections: Jobs and Skills Australia Employment Projections 2025–2035 (CC BY 4.0). The exposure index presented here is derived by FairWork Mate and is not endorsed by Jobs and Skills Australia or the Commonwealth of Australia.